#nobridge

  • 0 Posts
  • 6 Comments
Joined 1 year ago
cake
Cake day: March 14th, 2025

help-circle


  • When it comes to Nvidia GPUs the VRAM is the main thing to look for.
    For consumer cards it is:
    Entry level - RTX 5060 Ti 16GB RAM with a price point around 500-550 euro
    Mid - Buying a used RTX 3090 24GB RAM with a price point around 830 euro when I look at swedish second hand markets
    High - RTX 5090 32GB RAM with a price point around 3500 euro

    After that you end up looking at the RTX Pro Blackwell cards:
    Entry - RTX PRO 5000 Blackwell 48GB RAM ~5300 euro
    Mid - RTX PRO 6000 Blackwell 96GB RAM ~10100 euro

    It all depends on which models you want to run, you can definitely start playing around with Llama 3 8B and similar models with a 5060 Ti 16GB.

    If you’re looking at 24B-30B models you need the 24GB VRAM that RTX 3090 offers and get a larger context window if you go for the RTX 5090.

    If you’re looking to run Llama 3 70B then you need to go into the RTX Pro level of vram.

    All of this is based on running it with Nvidia cards, there’s also other setups such as Mac Studios with huge amount of RAM. They’re slower but allow for much larger models at the same price point.
    You could also run with AMD/Intel gpus but much software is built primarily for running CUDA (and Nvidia) gpus so it’s more work and not always compatible.

    I know you said no “monster rack” but I don’t really know what you classify as a monster. :)
    An ordinary gaming pc is also a good starter AI pc, so something like this allows you to do both:
    https://pcpartpicker.com/list/sFp4qd



  • When I wanted my current server to be available for more experimental loads I decided I needed a new machine and at that time all hdds had gone up in price, but a package of ds225+ with 2x16TB had not yet been price increased so I got the NAS “for free”.
    I would’ve repurposed an old gaming machine otherwise.
    Worth looking for those kind of “deals” where someone has forgotten to increase a bundle price.

    DS225+ with jellyfin and transcoding:

    I added 16GB RAM (Crucial 16GB (1x16GB) DDR4 3200MHz CL22 SODIMM) and the github.com/007revad/Transcode_for_x25 fix to enable proper transcoding for a docker install of Jellyfin and enough RAM to add other “production load” dockers in the future.
    Now I can play around all I want on my old server while Jellyfin and SMB shares remain available on redundant drives with automated backups. I’ve also blocked the NAS from the internet in my router now to ensure that Synology can’t “fix” the transcode fix.